Nvidia's Jensen Huang Targets 'Physical AI' as the Next Computing Frontier

Artificial intelligence will need to escape digital screens and learn the basic laws of physics if it is to reach its next major evolutionary phase, according to Nvidia Chief Executive Jensen Huang.

Speaking at the Hill and Valley Forum, Huang outlined a vision centered on "physical AI"—systems embodied in robotic machinery that can accurately perceive, reason about, and interact with the physical environment.

Beyond Digital Interfaces

While generative AI has spent the last several years mastering language and image processing within software environments, Huang argued that embodied machines require an entirely different cognitive layer. Instead of merely labeling objects in a video feed, physical agents must intuitively grasp physical dynamics: predicting trajectories, understanding how materials behave when dropped, and actively maneuvering around hazards in real time.

Bridging that gap will require massive synthetic environments. Nvidia has increasingly positioned its simulation software as the primary digital proving ground where autonomous machines can train against physical constraints millions of times before deploying into factories, hospitals, or public roads.

The Hardware Engine Behind Embodied AI

For Nvidia, this strategic focus extends well beyond high-level theory. While the company currently commands the data center market for training large language models, physical AI opens a second, potentially larger market: selling the specialized computing clusters needed to run real-time physics simulations alongside the edge processors installed directly inside machines.

The challenge, however, remains execution. Simulating edge cases with high enough fidelity to prevent fatal real-world errors has long bottlenecked the robotics industry. Whether Nvidia's simulation-first pipeline can accelerate physical autonomy as effectively as text generation remains the primary question hanging over the sector.